AI THESIS MONITOR
Blanc WM — Macro Tracker — Baseline 16 May 2026
● LIVE Refresh: Monthly
The AI capex cycle is bottlenecked by physical infrastructure — datacenter shells, electricity, transformers — not by GPUs. Sightline Climate reports only 5 GW of 16 GW announced for 2026 is in active construction. Transformer lead times have stretched from 50 weeks (2021) to 160 weeks (2026). This dashboard tracks whether the buildout matches forecasts, and alerts when the gap widens.

Five thesis pillars are tracked: (1) physical capacity buildout — actual GW commissioned vs announced; (2) hyperscaler capex discipline — MSFT/META/GOOGL/AMZN/ORCL capex actual vs guide and capex-to-revenue trends; (3) power & grid basket performance — equal-weighted GEV/ETN/PWR/CEG/VST/HUBB/NVT vs S&P 500; (4) credit stress — HY spreads, datacenter ABS issuance, bank offload activity; (5) compute pricing — GPU rental rates and inference token costs as leading glut indicators.

Composite Thesis Health Score (0–100) averages the five pillars. ≥70 green = thesis intact, stay invested; 40–70 amber = mixed signals, hold positions, do not add aggressively; <40 red = thesis broken, reduce power-infra basket, watch credit and hyperscaler capex cuts for confirmation.

Refresh cadence: first day of each month, 09:00 Zurich, via scripts/refresh_ai_thesis.py. Alerts: any pillar dropping >15 points month-over-month, or any individual metric breaching its threshold, fires a Telegram push to @blanc-wm. Baseline locked: 16 May 2026 — every metric tracks absolute level + delta vs baseline so trend remains visible across refreshes.

Sources: Sightline Climate quarterly DC outlook (paid, manual quarterly), Epoch AI Stargate tracker (free), SynMax satellite analytics (paid), SEC EDGAR 10-Q (auto-scrape, monthly), ICE BofA HY OAS via FRED API (free), PJM/ERCOT public APIs (free), Artificial Analysis token price API (free), Vast.ai/RunPod GPU rental scrape (free), Wood Mackenzie transformer lead time (paid, quarterly), Yahoo Finance API (basket pricing, daily).

58/ 100
Composite Thesis Health
AMBER — Mixed signals
Power & grid pillar leads (transformer scarcity confirms thesis), but capacity buildout is materially behind forecast. Hyperscaler capex still strong, FCF compressing. Credit stress contained but ORCL DC-debt offload bears watching. Action: hold power-infra positions, do not add new AI capex exposure aggressively until buildout pace clarifies.
Capacity Buildout
35
5/16 GW active · -45 delta
Capex Discipline
50
$345B run-rate · FCF -8pp
Power & Grid
80
160w lead time · basket +24%
Credit Stress
60
HY 295bp · $38B ORCL offload
Compute Pricing
65
Tokens -22% YoY · H100 $2.4/hr
Composite
58
5-pillar weighted average
⚡ Pillar 1 — Capacity Buildout (GW)
Announced
Active construction
Commissioned
💲 Pillar 2 — Hyperscaler Capex ($B, LTM)
2025 Actual
2026 Guide
2026 YTD run-rate
📈 Pillar 3 — Power & Grid Basket vs S&P 500 (YTD %)
Power Basket (eq-wt 7 names)
S&P 500
◆ Power Basket Constituents
TickerNameTierWeightLastPTUpsideYTD %SignalResearch
⚠ Stress Signals — Alert Thresholds
MetricCurrentThresholdStatus
🔔 Recent Alerts & Triggers
DateSignalAction
🔄 Refresh Mechanics & Alert Routing
ComponentSourceCadenceModeAlert Channel
Disclaimer. AI Thesis Monitor is an internal Blanc WM research tool. All thresholds, scores and forecasts represent informed estimates as of the baseline date and are subject to change as primary sources publish updates. Composite scoring methodology is illustrative; weightings have not been calibrated through formal back-testing. Power-infrastructure basket performance reflects an equal-weighted theoretical portfolio and does not include transaction costs or implementation slippage. Past performance does not guarantee future results. Not a recommendation to buy or sell any security. For individual client suitability, refer to the client-specific portfolio dashboard.